This study investigates the application of artificial neural networks (ANNs) for forecasting electricity usage in buildings, taking into account meteorological conditions and electricity pricing. Two feedforward artificial neural network models are utilized, employing the Levenberg-Marquardt and Quasi-Newton back propagation methods. The networks, designed with diverse depths and widths, are entirely interconnected. The primary objectives of the study are: (1) to develop and compare two predictive models, (2) to analyze the effects of various architectures, algorithms, and data processing techniques on model performance, and (3) to perform sensitivity analysis to evaluate the impact of input variables, including weather conditions. Empirical data from the Norwegian Maritime Museum underpins model training and validation. The results indicate that ANN models show potential in forecasting electricity usage trends, with performance differing according to the backpropagation algorithm, network architecture, and preprocessing techniques. Although the models accurately reflect consumption trends, they exhibit a deficiency in numerical precision, frequently diverging in absolute amounts. Sensitivity analysis identifies solar irradiance and electricity costs as key predictors, although ambient temperature significantly affects consumption during occupied hours. The results indicate that integrating meteorological and pricing information into predictive models can enhance the regulation of electricity consumption in structures. It is advisable to do additional study to incorporate supplementary factors and optimize ANN structures to improve predictive accuracy.